After a great week, lil_one bounces up to be tied for first with lunatica. However, 5 people are within 5 pts of the leaders. https://sites.google.com/view/nwsl/12 For this week I added the game list with individual predictions. When I have a moment in the next few days, I'll look week by week to double check the 1-2 pt differences in total scores just to confirm.
I did that work for you: basically, we were carrying on some slight mistakes since the very first weeks, and the bottom line is that your scorings are correct (as one could expect, since they had been automatically calculated; but you could have anyway made some little mistakes in inserting data of single predictions, as you had done with 59Amerinorsk, so it was worth checking to make sure that we could rely on current rankings). Details: - there were 5 situations where the pointings made by STT weren't matching with the ones Lucy had made with her Excel program: @holden had one more point in STT's rankings than in Lucy's: this remounts to first week; for some reason, Holden had received 2 points for his prediction of SEA 1 NJ 2; the match had finished 1-1 instead, so he should have had just the single point for correctly predicting the number of goals scored by Seattle. @CoachJon had one more point in STT's rankings than in Lucy's: the mistake was made in pointing the same match (we had seen in the past STT making multiple mistakes on a single match ), although for different reasons, I guess. Coachjon had predicted 2-0, so he should have got 0 points. If he got 1 instead, it's because during first two weeks STT was adding 1 point for predicting the correct total number of goals scored in the match (and 2-0 have the same number of goals than 1-1). Later, STT removed from the rankings all of the points gained this way (see post #80 of this thread, page 4), but in that circumstance he forgot about CoachJon, who thus mantained his ill-gotten gain until Lucy-fun police came in. @LucyFearsTheMorningStar had one point less in STT's rankings than in Lucy's: once again, it happened in first week; the prediction HOU 3 CHI 2 was credited 0 points, while Lucy should have got 1 for at least correctly nailing Houston's win (by 2-0) @Blaze20 had one point less in STT's rankings than in Lucy's: this happened in second week instead (probably, in the first two weeks, we weren't used to double-checking our personal scores yet, and this didn't help STT with his work: later, we started spotting single mistakes and pointing them to STT, who could subsequently correct them in the following week). Blaze was given just one point for predicting CHI 2 KC 1 (real result: 1-0), while she should have had 2 (1 point for predicting Chicago's win + 1 point for correct goal difference) @Smallchief had two more points in STT's rankings than in Lucy's: this one was really tricky! I checked and double-checked multiple times, but I couldn't find the mistake to save my life! Then it dawned on me: I had picked STT's totals from weeks 7+8 as my starting point: well, if you check, in that circumstance when he had pointed two weeks in a single post, STT had made the following addition: 28 (Smallchief's starting points) + 10 + 4 = 44 (instead of 42!) Thus, the two "ghost points". I guess right now, we can go on peacefully with Lucy's chart.
For what it's worth: As a test, I took a look at the numbers of weeks for which our predictors have made predictions as compared to their point scores. I made an assumption that a predictor did not make any predictions if the score for the week was 0. My reason for taking a look at this is that I'm interested in how strictly data-based predictions compare to human predictions. I've tried to make my predictions as data-based as possible, regardless of my personal views, so I'm treating my predictions as strictly data-based although they actually require some limited and constrained judgments on my part as I described in an earlier post. The results of my look also are skewed, as I've been making predictions only for the last 5 weeks, which should make my average points per week higher than those of predictors who have been posting since the beginning of the season (when, at least for a while, I would have had to use data from last year). In addition, predictors may not have made predictions for all games in a week for which they made some predictions, and I haven't taken that into account. Nevertheless, I still find the results interesting: cpthomas (data driven) 7.6 points per week lunatica 6.6 lil-one 6.6 blaze20 6.3 LFTMS 6.3 blissett 6.2 bluecrimson 6.2 CoachJon 6.2 holden 5.6 59Amerinorsk 5.3 smallchief 5.2 SiberianTT 4.5 Since I only have done the last 5 weeks, which may make the above comparison "unfair" here are the comparisons using only those 5 weeks and limited to predictors who made predictions for all of those weeks: CoachJon 7.9 per week cpthomas 7.8 LFTMS 7.0 bluecrimson 7.0 lil-one 6.8 lunatica 6.2 blissett 5.8 blaze20 5.6 smallchief 4.8 This is more in line with my experience, which though limited is that data driven predictions will do better than most of the humans, but that some humans will do better than the data driven predictions.
So what you're saying is we should wait for your predictions and then copy you with maybe 1 or 2 changes? I will add week 11 is somewhat scewed for some. Smallchief only got the 4 games Wednesday in and blaze20 and lil_one each only got the 5 weekend games.
Actually, I think a wiser course would be to look at the most recent data table before making predictions, to see what you would say if you relied only on the data. Then, add any additional perspective you think is appropriate to see where you would modify what you think the data suggest. For example, some of the added perspective I can think of would come from the injury list, your sense of how teams are trending, international call ups, and roster changes. Those all are things that my rules for my predictions don't allow me to consider.
There was a very popular weekly contest in Italy, based on predictions of the results of Italian club league. My dad used to always play and he had an approach very similar to yours: historical data + most of the additional info you mention. He often had quite good result, on average, but... he never won!
Not sure how much of a factor it is, but, on 12 weeks only, even a single week of this kind could have his weight (for instance: in week 1, I predicted just one game out of 5; I nailed and made 3 points, but it could have been much worse. Or much better, had I predicted the whole matchday). There are multiple cases like this, if I remember well (Lucy mentioned two of them).
Predictions for this coming weekend: Portland 1 v North Carolina 1 NJNY 1 v Chicago 1 Orlando 2 v Kansas City 1 Houston 2 v Washington 1 Seattle 3 v Boston 1
Updated Tables. lunatica and lil_one remain neck and neck. No one got Sea-Bos. Only 1 person got the outcome of Por-NC right while Hou-Was had 5 get the exact score right.
I am trying to fight to remain at least close to the top of the rankings, but it becomes harder and harder, if Seattle doesn't help me by making results!!! P.S. It's a shame BlueCrimson skipped this week. I don't like beating good opponents with just continuity.
Predictions for 7/22 games: Chicago 2 v Orlando 1 Kansas City 1 v North Carolina 1 Houston 1 v Boston 0 Seattle 2 v NJNY 1 Portland 2 v Washington 0 Interesting data tidbit: So far this season Home wins 37 (52.9%) Ties 17 (24.3%) Home Losses 16 (22.9%)